Animal experiments were approved by Iowa State University Institutional Animal Care and Use Committee, log #19-211. Animal distress was minimized during experimental procedures via enrichments (ex: perches). Open-floor pens were implemented to enable social interactions between chickens, and chickens were fed with no-medicated feed and water ad libitum. Upon experiment completion at 5 days post pen-placement, chickens were euthanized via CO₂ asphyxiation in accordance with American Veterinary Medical Association Guidelines (2020).

Chickens used in video recordings for the creation of VR were enrolled from and maintained on three privately owned backyard flocks in Iowa during summer 2019. These flocks consisted of mature hens, roosters, pullets, and chicks, including a diversity of fancier-type egg-laying, meat and dual-purpose breeds. Housing included one or more indoor facilities, equipped with perches, nest boxes, feeders, drinkers and litter areas, with access to an outdoor fenced scratch area or unfenced open prairie with grasses, shrubs and flowers. Chickens used in the laboratory testing of VR consisted of 34 White Leghorn hens (Dekalb Whites) at early-lay stage (15 weeks post-hatch), randomly split into two groups, one exposed to virtual reality (VR) and the other unexposed (CON) (n = 15) via staff blinded to the experimental groups. Birds were individually identified using food dye on plumage and colored leg bands. Each group was housed in a 9 ft × 9 ft × 7 ft pen with steel spindle walls and pine shavings litter, and pens were maintained at 74°F. In both VR and CON rooms, pen walls were wrapped with vinyl projector screen material on all four sides. Each pen was equipped with a feed hopper, drinker, perches, nest boxes, and a dustbathing area. The photoperiod was 16L:8D. Husbandry consisted of daily inspection of the chickens, replenishing water and a commercial layer diet was provided ad libitum.

Hens from the VR and control groups were housed in 9 ft × 9 ft × 7 ft steel cages wrapped with vinyl projector screen material on all four sides. In the VR pen (Figure 1), video-recorded scenes of free-range chickens (see Supplemental Materials for example videos) were projected on each of the four walls during daylight hours. Each wall depicted a free-range flock interacting with one of the following scenes: 1, perching area; 2, nesting area; 3, outdoor dirt yard scratch area; 4, outdoor vegetative area. The storyboard for wall was constructed with hourly segments depicting scenes intended to provoke desirable behaviors (e.g., nesting, dustbathing, preening, perching) appropriate to the known ethogram and time budget of the laying hen. For example, VR projected behind the nest boxes displayed hens actively performing nesting behavior during the morning hours, when oviposition typically occurs, whereas empty nests with no hens present were displayed during afternoon and evening hours when oviposition is unlikely. The perching area VR storyboard switched hourly between still images of perches with no hens present, followed by video images of hens actively perching during the daylight hours, and concluding with a still image of perches with motionless hens perching immediately prior to and during the dimming lights and dark periods.

Prior to necropsy, all hens were weighted and evaluated according to welfare characteristics, including plumage damage, cleanliness, keel bone deformation, comb pecking wounds and comb abnormality as previously described (Oliveira et al., 2019).

Prior to euthanasia at day 5 of VR or control treatment, fresh blood (100 µL) was collected via wing vein via filter-sterilized heparin (1,000 U/ml)-coated needles and immediately placed in citrate-coated tubes at room temperature for 4 h. Blood was centrifuged at 4,000×g for 20 min, and non-hemolyzed serum was individually stored at −80°C. For quantification of corticosterone, serum samples were run in triplicate using a commercial corticosterone ELISA kit (Enzo Life Sciences) in accordance to the manufacturer directions.

Prior to euthanasia at day 5 of VR or control treatment, fresh blood (200 µL) was collected from the right-wing vein of hens via filter-sterilized heparin-coated needles and placed on ice. Blood was evenly split into 2 pools per treatment group (n = 7 hens per pool), then diluted 1:8 in CO₂-independent media with 2 mM L-glutamine (Gibco). Cell viability was determined to be greater than 98% via trypan blue staining for all pools. Prior to experiments, reference E. coli strains (Table 1) were stored in peptone-glycerol solution at -80°C. Strains were grown in Luria-Bertani (LB) agar (0.1% glucose) overnight at 37°C, colonies were mixed into phosphate-buffered saline (PBS) until OD₆₀₀ reached 0.1, and solutions were diluted to reach 10² CFU/20 µL. Individual wells in 96-well plates were filled with whole blood solution and bacterial inoculum (9:1, respectively) and incubated for 30 min at 40°C. After brief resuspension, samples were serially diluted, plated on MacConkey agar, and incubated overnight at 37°C. Assays were performed in triplicate.

To calculate H:L ratios, a biomarker for stress (Gross and Siegel, 1983), approximately 3 μL of fresh blood (acquired prior to euthanasia as described earlier) were aliquoted and smeared onto a standard microscope slide (one slide per hen, n = 15 per group), briefly air dried, fixed and stained using a Neat Stain Hematology Staining Kit (Polysciences, Inc.). For each slide, approximately 200 leukocytes were enumerated via light microscopy based on phenotypic characteristics (de Macchi et al., 2013). Thereafter, H:L ratios were then calculated by averaging heterophils and lymphocytes identified within each slide.

Post-euthanasia at day 5 of VR or control treatment, intestinal segments (i.e., ileum and ceca) were aseptically removed from birds in both groups (n = 15 per group). Ileal and cecal contents were squeezed from the intestinal segments into respective tubes. Mucus from each section was collected by longitudinally dissecting intestinal tissues to expose the intestinal lumen and then gently scraping the mucosa with fresh, autoclaved glass microscope slides, which were then directly transferred into respective cryotubes. The remaining tissues were then placed into respective cryotubes. All sample cryotubes were briefly homogenized with 0.2 M perchloric acid (1:10 sample-acid ratio) and flash-frozen via liquid nitrogen prior to −80°C storage. For neurochemical and short chain fatty acid analyses, samples were homogenized using bead mill homogenization (Omni Bead Ruptor Elite) and centrifuged at 2,950 × g at 4°C. Thereafter, supernatant was transferred to 2–3 kDa spin filter and centrifuged again at 2,950 × g at 4°C. Flow-through was collected and then analyzed via ultra-high-performance liquid chromatography with electrochemical detection (UHPLC-ED) using the 99 Dionex UltiMate 3000 with MD-TM Mobile Phase Solution as sample diluent (Fisher Scientific) as described previously (Villageliú et al., 2018b).

Total DNA was isolated from 0.25 g intestinal samples (ileal mucus, ceca mucus, ceca content) via bead-beating using the DNeasy PowerSoil Kit (Qiagen). Extracted DNAs were assessed for quality using a NanoDrop 2000 spectrophotometer 260–280 nm ratios. Concentrations were determined using a Qubit fluorometer with the double-stranded DNA broad range kit (ThermoFisher Scientific) and adjusted to 50 ng/μL in nuclease-free water. Ileal content was initially-sampled, but insufficient DNA yield prevented further analyses. Library preparation was performed using the MiSeq and HiSeq2500 kit (Illumina) following all manufacturer’s instructions with 151 × 151 paired-end MiSeq sequencing (Illumina). DNAs were then sequenced via Illumina MiSeq (v3) at the Iowa State DNA facility. For 16S rRNA analysis, QIIME2 (version 2019.10) was used to analyze the 16S rRNA data between all sequenced groups. Sequences were demultiplexed using the QIIME2 demux emp-paired function and denoised using the QIIME2 plugin DADA2. The number of good quality reads for taxonomic assignment ranged from 37,495 to 114,115 reads. GreenGenes database (version 13.8) at the 99% operational taxonomic units (OTUs) spanning the V4 and V5 regions (515F, GTGYCAGCMGCCGCGGTAA; 926R, CCGYCAATTYMTTTRAGTTT) was used to classify each of the reads using QIIME2’s feature-classifier function. Alpha and beta diversity analyses were calculated using QIIME2’s built-in functions. Mmvec plugin (Morton et al., 2019) was used to explore taxonomic balances and taxonomic group differences between treatment groups. T tests were used to determine statistical significance between groups (see Section 2.10 for further details).

GraphPad Prism software version 6.0 (San Diego, CA) was used to calculate significance and construct figures for all analyses. Parametric, unpaired t-tests (two-tailed; α = 0.05) were used to compare differences between CON and VR groups. For bactericidal assays, where samples were pooled, assays were performed in triplicate per pool (n = 6 per treatment group). For 16S rRNA bacterial abundances and neurochemical data, differences were compared within sample site (ex: content, mucus, or tissue from ceca or ileum). For all statistical tests, data sets for each group were assumed to follow a normal distribution. p values <0.05 were considered significant.

General welfare assessment has shown no significant differences in body weight between control (1.15 kg ± 0.066) and VR (1.15 kg ± 0.063) groups (Supplementary Figure S1). Additionally, birds in both treatment groups exhibited no signs of injury from negative behaviors like pecking.

Using a blood bactericidal assay to test resistance to avian pathogenic E. coli (APEC), we found that blood from VR hens exhibited higher APEC killing to the three APEC serotypes O1, O2, and O78 tested compared to CON hens (p < 0.05; Figure 2A). This phenotype was positively-associated with an increase in serum corticosterone levels (p < 0.01; Figure 2B). However, H:L ratio was not significantly different between treatment groups (p > 0.05; Figure 2C).

Ultra-high pressure liquid chromatography (U-HPLC) data for neurochemicals and related metabolites in the content, mucus, and tissue of the chicken gut are summarized in Figure 3 (ileum) and Figure 4 (ceca). In the ileum, all detected neurochemicals were not significantly different in each site except for salsolinol (Figure 3C), which was higher in the content of CON hens (p < 0.0001) yet higher in mucus of VR hens (p < 0.01). Similarly, only salsolinol levels changed in the ceca between groups (Figure 4E), although differences were only seen in the tissue with VR hens having greater salsolinol levels (p < 0.05).

Using 16S rRNA sequencing and QIIME2 analysis, microbiome alpha diversity (Faith PD, Pielou’s Evenness) was similar between groups (Supplementary Figure S2). Similarly, microbiome beta diversity was highly similar between groups at each intestinal site, although samples from the ileum mucus clustered distinctly along Axis 1 away from other intestinal samples in every beta diversity measure (Figures 5A–C, Weighted Unifrac; 5D-F, Unweighted UniFrac; 6A-C, Jaccard; 6D-E, Bray-Curtis). Looking at bacterial abundances, several taxa were consistently lower in VR hens compared to CON. For example, looking at ceca genera, Megamonas were lower in the mucus and content (p < 0.01), whereas Ruminococcus and Slackia were lower in content alone (p < 0.05; Figure 6A). Looking at the species level, Barnesiella viscericola and Blautia producta were reduced in the ceca mucus of VR hens (p < 0.05; Figure 6B). Uniquely, Mucispirillum schaedleri was the only taxon higher in VR hens, being greater in ileum mucus versus CON hens (p < 0.05; Figure 7B). Notably, these shifts did not have any impacts on short chain fatty acid (SCFA) production (Supplementary Figure S3).

To explore the possibility that specific bacterial taxa may be related to intestinal neurochemical synthesis, we used the QIIME2 plugin mmvec (Morton et al., 2019) to draw relationships between neurochemical and microbiome data from the ileum mucus, ceca mucus, and ceca content. Using an Emperor plot to display microbial or metabolite co-occurrences, two phyla co-occurrences (1: Firmicutes and Cyanobacteria; 2: Tenericutes, Deferribacteres, and Euryarchaeota) appear frequently along Axis 1 (Supplementary Figure S4A). These co-occurrences are amplified when plotted against Axis 3 (Supplementary Figure S4B) versus Axis 2 (Supplementary Figure S4A). Similarly, norepinephrine and homovillanic acid appear to frequently co-occur regardless of axes (Supplementary Figure S4). Lastly, to observe microbial-metabolite co-occurrences, inferred log conditional probabilities (LCPs) of microbial correlations with neurochemicals are displayed as a heatmap in Figure 8 and numerically summarized in Supplementary Table S1, in which positive probabilities are distinguished as red whereas low and negative values (i.e., light red to dark blue) indicate no relationship, not necessarily a negative relationship (see Github mmvec page: https://github.com/biocore/mmvec). L-DOPA and norepinephrine are positively related to Cyanobacteria (LCP = 1.52) and Euryarchaeota (LCP = 1.78), respectively. Dopamine and its waste metabolite salsolinol are positively related with Deferribacteres (LCP = 1.51 and 1.73, respectively) and Tenericutes (LCP = 1.69 and 1.71, respectively). Serotonin (5-HT) and its waste metabolite 5-hydroxyindoleacetic acid (5-HIAA) are positively related with Bacteroidetes (LCP = 3.19 and 2.05, respectively), Cyanobacteria (LCP = 3.43 and 2.33, respectively), Firmicutes (LCP = 3.30 and 2.63, respectively), and Proteobacteria (LCP = 2.74 and 2.07, respectively), with serotonin being specifically related with Spirochaetes (LCP = 1.71) and Tenericutes (LCP = 1.72). Uniquely, homovillanic acid and DOPAC are not related to any bacterial taxa at the phylum level.